DanielLeens commented on PR #12216: URL: https://github.com/apache/seatunnel/pull/12216#issuecomment-5617133892
Reworked the fix in `68c0185b0970fd6b337d5c64fb8cd38d2fc410c1` after tracing the previous CI failures. The prior head was not valid: its UUID was created once per per-table `AssertSink`, so multi-table writers did not share table-name state and E2E jobs failed with errors such as `[people] is not equal to [people, companies]`. The same file also contained an embedded NUL, which made both Windows unit jobs fail Spotless. The current head now follows #12116's confirmed writer-local row-count contract: - `LONG_ACCUMULATOR` is replaced by a non-static map owned by each `AssertSinkWriter`. - UUID plumbing and constructor changes are removed. - `TABLE_NAMES` behavior is restored unchanged and remains explicitly out of scope pending its separate contract decision. - The regression now deterministically proves two same-table writers cannot affect each other's MAX_ROW result. Module Spotless and `git diff --check` pass. The effective PR diff is two files with no public API or serialization change; fresh CI has started from this head. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
